Classical stochastic models for chemical reaction networks are given by continuous time Markov chains. Methods for characterizing these models will be reviewed focusing primarily on obtaining the models as solutions of stochastic equations. The relationship between these equations and standard simulation methods will be described. The primary focus of the talk will be on employing stochastic analytic methods for these equations to understand the multiscale nature of complex networks and to exploit the multiscale properties to simplify the network models.